Simulation and modelling of transport networks and ... · Simulation and modelling of transport ......
Transcript of Simulation and modelling of transport networks and ... · Simulation and modelling of transport ......
22th October, 2012
Essen, Germany
TU Dortmund, Institute of Transport Logistics
26th European Simulation and Modelling Conference 2012
Simulation and modelling of transport networks and logistic nodes
Prof. Dr.-Ing. U. Clausen, Fraunhofer-Institute for Material Flow and Logistics
TU Dortmund, Institute of Transport Logistics
Dipl.-Logist. D. Diekmann, Dipl.-Logist. I. Goedicke, Dipl.-Logist. J. Kaffka, Dr. math. F. Meier, Dipl.-Geogr. C. Thaller
ESM 2012, 22th October, 2012
Overview
2
Challenges in transport logistics
Approaches for modelling and simulation
Conclusion & Outlook
ESM 2012, 22th October, 2012
Challenges of the future
3
Individual mobility and individual
provision of goods and services
Robust and secure logistics solutions for
conurbations
Environmentally sound and resource-
saving logistics processes
Logistics has to make a substantial contribution to a positive development of the present societal challenges.
Protecting the environment and preser- ving resources
Securing the supply of urban systems
Preserving individuality
Increasing scarcity of raw materials
Limited absor-bance capacity of ecosystems
Urbanization Demographic
change Growing need for
security
Diversified consumption behavior
Various leisure activities
ESM 2012, 22th October, 2012
Welcome to Ruhr Logistics Metropolis
4
Strength of Ruhr Metropolis Europe‘s third-largest
urban centre after Paris and London
Turntable for the national and European traffic
5,700 logistics firms along the entire value chain
160,000 employed in the logistics sector and logistics-related industry
Rapid growth of logistics sector
As one of Europe’s leading logistics locations the Ruhr Metropolis develops, implements and exports logistics solutions to meet the societal requirements.
http://www.ruhrbarone.de
http://www.ruhrbarone.de
ESM 2012, 22th October, 2012
EffizienzCluster LogistikRuhr
5
One of 15 Leading-Edge Clusters within the High-Tech Strategy 2020
Partners 120 companies 11 education and research
facilities Several intermediate actors
Declared strategic goal To solve conflicts between
efficiency and individuality To create tomorrow’s
individuality with 75% of today’s resources
Dortmund / Unna
The close collaboration between science and economy within the EffizienzCluster assures a transition of innovative ideas into marketable logistics solutions.
ESM 2012, 22th October, 2012
Challenges in transport networks & objectives of optimizations
6
significantly increasing transport performance, shipment volumes and GHG-emissions
(new) restrictions for pickup and delivery tours in urban areas
pollution by common vehicles
increasing complexity and diversification in customer demands
time windows, late pick ups, early deliveries
Methods to enable efficient and sustainable transport networks under the given challenges and restrictions.
0
200
400
600
800
1000
road
inland waterway
rail
bn. tkm
Shell Lkw-Studie
ESM 2012, 22th October, 2012
System modelling
7
model =
simplified, abstract copy of an existing or virtual system which represents the essential elements and interactions of the system
model classifications
purpose forecasting
simulation
decision, optimization
type of information quantitative
qualitative
type of abstraction deterministic – stochastic
static – dynamic
total – partial
ESM 2012, 22th October, 2012
Comparison of methods
static models can be build more easily dynamic models are more
complex to build but more realistic
8
paper & pencil
spreadsheet
linear programming
queueing theory
simulation
dynamic & complex behaviour
system size
big systems can only be handled with the help of static modelling
logistics systems are
highly dynamic
Figure: Incontrol Simulation Solutions
ESM 2012, 22th October, 2012
Overview
9
Challenges in transport logistics
Approaches for modelling and simulation
Conclusion & Outlook
ESM 2012, 22th October, 2012
Insights into tranport logistics modelling and research – overview
Economic Modelling
Operations Research Simulation
10
ESM 2012, 22th October, 2012
Integration of delivery time windows in urban areas structures, activities and
behaviour of forwarders
Results for the integration of delivery time windows in the Ruhr Area e.g. increasing amount of
time windows:
Static modelling of commercial traffic
11
It is essential to integrate logistic parameters and effects to determine actually induced freight flows and stress for infrastructure.
180
2005
160
140
120
100
0 2007 2003 2001 1999 1997 1995
GDP transport performance transport volume
transport distance
+16% GDP and other aggregate data
are an insufficient basis for deriving freight transport
Logistics structures and strategies have to be considered
Road freight transport & GDP in Germany
inde
x 19
95 =
100
Destatis 2007, Iddink 2009, Seredenko 2008/2011
+4%
# delivery tours
ESM 2012, 22th October, 2012 12
Commercial Traffic Modelling – Integration of logistics nodes
Integrate characteristics and specific throughput volume of logistics nodes in the demand modelling of freight transport Typology of logistics nodes
Empirical analysis of different hub types
Objective: new demand theory for the integration of transport nodes in transport models
Source: DLR, ITL 2012
Which models can help to predict the operations of transport logistics?
Airport
4% 3%
42% Road
Further Sea port
13%
Inland port
7% Freight villages
11% Intermodal
11%
CEP
4%
Rail
5%
ESM 2012, 22th October, 2012
Mathematical Optimization – Improving the efficiency of transport networks
13
Problem: Transport small shipping volumes between a lot of dispatchers and receivers. Ad hoc solution: A direct transport network. Disadvantages: Low capacity utilization Large number of connections to
maintain.
Better solution: Consolidate and deconsolidate goods at hubs:
Use economies of scale.
Hub location problem: Where should we build hubs to minimize costs while fulfilling all real world conditions?
ESM 2012, 22th October, 2012
Hub Location Problems – Building a model
14
1. Model the important
restrictions/conditions
2. Model the costs
Maximal capacity of hubs and/or connections
Maximal total time for each point-to-point connection or for complete routes
Allow any route from A to B or make restrictions (e.g. every dispatcher of country X delivers to hub H)
Fixed and variable transport costs (depend on the chosen routes and the capacity utilization)
Costs for building and maintaining the hubs and undertaking transshipments
ESM 2012, 22th October, 2012
Hub location challenges
15
Real world conditions Example Computability?
Few routes allowed Postal services require sending letters for sorting to nearest hub
Computable, but little improvement to ad hoc solution
A lot of feasible routes In LTL you might allow to route every shipping volume individually.
Huge solution space which (usually) contains very good solutions. Problem: You often overlook them.
Real world conditions Example Computability?
Costs depend on kg/m³ on each route (rare case) Computable linear model
Costs depend mostly on the number and type of vehicles on each route
LTL, air (cargo and passengers)
A lot of integer variables. Difficult.
Routing
Transport costs
ESM 2012, 22th October, 2012
Hub Location Problems – Research objectives & approaches
LTL Problem Measure transport costs by vehicle and allow a lot of routes (more realistic) How do we solve this? (CPLEX and GUROBI fail for 50 dispatchers or more)
Approach Combine the following three elements: Standard branch-and-cut techniques (in combination with commercial solvers) Problem-specific heuristics: They create good feasible solutions and measure the
“quality” of the possible hubs. Quadratic techniques: Additional restrictions can be derived mathematically from the
quadratic nature of the cost function. This allows to reduce the search space.
Preliminary results are promising Good (though not optimal) solutions can be computed for 60 dispatchers in 12
hours.
16
Robustness of the solution under stochastic influences, especially concerning the network nodes cannot be evaluated with the approach.
ESM 2012, 22th October, 2012
High level of system dynamics High level of complexity
Modelling of nodes in logistics networks
17
Significant peak hours in system loads Combination of manual and automated handling
processes Heterogenous goods
Challenges for terminal operators
ESM 2012, 22th October, 2012
Using simulation to control system dynamics and complexity
18
Layout planning
Operating strategies
Yard management
Door as interface
Terminal building
Size and utilization of external areas
Loading & unloading locations; assigning destinations to doors
Personnel management;
loading concepts Allocation rules
Management of traffic flow &
shunting; crane control
Size and utilization of internal areas
ESM 2012, 22th October, 2012
Modelling of logistics nodes
The used software …
provides typical modules of logistics nodes for microscopic modelling allows implementation of specific characteristics by individual programming
19
Syst
em
com
pone
nts
Static: doors, buffer areas, gate-keeper, roads …
Dynamic: forklifts, vehicles, trains, containers…
Setti
ngs
Driving speed (depending on vehicle type)
Process times, capacities
…
Rea
listic
pro
cess
ch
ains
Process steps of different vehicles & sending types
Controlling strate-gies of resources (forklifts, cranes …)
Realistic scale modeling of complex systems results of the simulation experiments can be transferred to the real system
ESM 2012, 22th October, 2012 20
Reference project – Dispatching rules for forklifts
in-house (building)
component process functional unit
arrival departure
waiting time
waiting time
shunting shunting
unloading loading
sorting and buffering
external (yard)
Interface yard building
gatekeeper / barrier system
waiting area / parking
shunting vehicle
door
manual or automated sorting
Research objective Develop dispatching rules and evaluate their
impact on terminal activities with the help of simulation
Procedure Mapping system load data, processes,
operating strategies in forwarding agencies Development and implementation of different
dispatching rules (single and multiple attribute rules)
Creation of the model in the simulation environment and simulation study
Picture: Jungheinrich
ESM 2012, 22th October, 2012
Reference project – Dispatching rules for forklifts
Forklift initiated rules Longest Waiting Time (LWT) Shortest Travel Distance Pickup
Point (STDPP) Maximum Queue Size (MaxQS)
21
Assignment not possible -forklift waits for order
Transport task-initiated dispatching rules Status not possible
Assignment not possible - order waits for free forklift
No choice assignment Forklift-initiated dispatching rules
No assignment required
Number of waiting tasks0 1 >1
Number of free
forklifts
0
1
>1System change created
by a new task
System change created by an empty forklift
Combination “First Transporter First-Longest Waiting Time” is considered as basic scenario. The dispatching rules LWT, STDPP, FTF and NV are tested with all possible combinations.
Task-initiated rules First Transporter First (FTF) Nearest Vehicle (NV) Least Utilized Vehicle (LUV)
ESM 2012, 22th October, 2012
Dispatching rules for forklifts – Simulation model
22
Key figures
Ground space of 6.300m²
24 unloading / 36 loading doors
4.300 handling units per day
23 national and 33 regional destinations doors
buffer areas internal transport network
fork lift fleet
ESM 2012, 22th October, 2012
Dispatching rules for forklifts – Simulation model
23
Key figures
Ground space of 6.300m²
24 unloading / 36 loading doors
4.300 handling units per day
23 national and 33 regional destinations
ESM 2012, 22th October, 2012
Dispatching rules for forklifts – Results
Average full travel time: Constant in the different scenarios
Average empty travel time : 13% faster by changing forklift
initiated rule “Longest Waiting Time” to “Shortest Travel Distance Pickup Point”
11% faster by changing task initiated rule from “First Transporter First” to “Nearest Vehicle”
26% faster by changing both rules (FTF-LWT to NV-STDPP)
24
Tim
e [s
]
0
5
10
15
20
25
30-26%-11%-13%
NV-STDPPNV-LWTFTF-STDPPFTF-LWT
Simulation gives a unique opportunity to test different strategies and evaluate the potential to increase system efficiency.
ESM 2012, 22th October, 2012
Conclusion Travel path optimized forklift fleet control reduces
the empty travel time of the forklifts but requires process changes and additional investments
Scanning of shipments after unloading and tracking positions of forklift trucks is necessary
Outlook Extension of the single to multi attribute strategies
for the fleet control Apply strategies for more complex terminal shapes
Dispatching rules for forklifts – Conclusion and Outlook
25
Terminal operations involve a lot of decisions (e.g., door assignment, vehicle priorization, resource allocation) – what is the best strategy mix?
ESM 2012, 22th October, 2012
Research project – Statistics and Simulation
26
Research objective Develop a method to identify the best possible
mix of operating strategies in forwarding agencies
Statistic based experimental planning
Procedure Mapping system load data, processes,
operating strategies in forwarding agencies Generate input data on a kernel density
based approach Develop evaluation models and logistical
hypothesis Identify the best mix of operating strategies
with loops of experimental planning, simulation studies and recursive partitioning
ESM 2012, 22th October, 2012
Modelling external systems – Container terminals
95% of goods worldwide are transported with ships
Key factor in maritime transport is the container
Shipping volume of containers grew by the factor 5 in the past 20 years
27
TRIMODAL
BIMODAL
YARD
Full Container (Import/Export)
Warehouse and stuffing
Empty Container Depot
Truck Handling Area
Train Handling Area
Ship Handling Area
Böse 2004, Port of Hamburg 2012
Container terminal material flow system Container throughput of the Port of Hamburg
Increase of throughput in terminals but space to handle container is limited
Container terminals are complex material flow systems with many subsystems
Large impact of stochastic influences, interdependencies of subsystems and complex decisions indicate advantages of simulation methods.
ESM 2012, 22th October, 2012 28
Research project – Managing a Trimodal Hinterland Hub
Research objective Merging of mathematical optimization
and simulation in one tool Support in planning and operating
new intermodal terminals Meet challenges of inland port
terminals
Methodologies Automated layout creation for
modelling, intelligent operating strategies
Preterm optimization of resource allocations
Crane control and sequencing of container movements using mathematical optimization
ESM 2012, 22th October, 2012
Different scenarios can be analysed with the help of the developed simulation suite
Online Optimization: every change of state of the system induces recalculation
Strategies for optimal terminal operations
29
Global controlling strategies Layout decisions Vehicle loading point allocation Human resource planning
Adh
eren
ce to
sch
edul
es %
Stock [TEU]
Mea
n cy
cle
time
[t]
Capacity100
80
60
40
20
max
Cycle time
Adherence to schedules
Utilization
Nyhuis and Wiendahl 2002
Operational strategies For handling equipment (e.g. crane
control) Priority parameter for every
company and market requirement Importance can be adjusted by the
weighting per parameter
]1,01
[ Container; ∈∈⋅= ∑=
ii
n
iij pjgpP
ESM 2012, 22th October, 2012
Strategies for optimal terminal operations – case study
Inland port container terminal 3 handling cranes 1.000 Ground slots 4 loading tracks for trains 400 meter quay for 2 barges 20 loading points for trucks
Increased terminal performance by optimized layout and control
strategy (priority number)
30
Current layout of the observed container terminal
-3.0% +4.0% -2.5% -6.0%
full travel ratio of cranes
truck cycle time
handling factor
waiting time at loading
point
ESM 2012, 22th October, 2012
Overview
31
Challenges in transport logistics
Approaches for modelling and simulation
Conclusion & Outlook
ESM 2012, 22th October, 2012
Conclusion & Outlook
Modelling characteristics are determined by System specifications
system load, complexity of system interactions, system dynamics … Scope of analysis
robustness, search for optimum, comparing operating scenarios …
Combining the advantages of different methods is a main objective but very difficult to achieve
32
ESM 2012, 22th October, 2012 33
Thank you for your attention Prof. Dr.-Ing Uwe Clausen TU Dortmund, Faculty Mechanical Engineering Institute of Transport Logistics Leonhard-Euler-Str. 2, 44227 Dortmund, Germany Tel.: +49 231 755-6335 E-Mail: [email protected]